Image File Formats and Organization
Image File Formats and Organization |
|
Richard H. Wiggins, III, MD has reported no financial interest, arrangement or affiliation with a commercial organization that may have a direct or indirect influence in the subject matter of this presentation.
Objectives
Contents
Picture Archiving and Communication Systems (PACS) have greatly assisted with the conversion from hard copy teaching files into the digital environment. As this transition progresses, an increasing number of radiologists are using the Internet as a vehicle for dissemination of teaching case material, usually with HyperText Markup Language (HTML) or Extensible Markup Language (XML) [1-3]. There are many, very different image file format types, and it can be a daunting task to choose the ideal one for a particular Digital Teaching File (DTF) project. The correct image file type for DTF use depends on many factors. This may be simplified for radiologists with a PACS with limited abilities for exporting images. The steps involved in getting an image into digital form (acquisition, storage, manipulation, and presentation) must be taken into account when planning such a project. Factors such as spatial and contrast resolution, as well as compression techniques can also play an important role in terms of storage space required and transmission times of images, especially over relatively slow bandwidth networks.
The two main forms of digital image file formats are the vector-based and bitmapped graphics, each of which has different strengths and weaknesses. There are also image file formats referred to as "metafiles" that can accommodate both bitmap and vector data. The individual image file format types also have differences, which must be considered along with the data structure and organization prior to creating a DTF. The most common image file formats are discussed below with others listed afterward with brief definitions.
Vector-based, or object-oriented, image files have in the past been primarily utilized for illustrations or drawings, which are less complicated than photographic images. The increasing popularity of these types of image files has grown considerably over the past year, in part due to the flexibility and interactivity possible in Web development. These file formats store data as a series of instructions on how to create the graphic file. These instructions usually take up less space than saving information about every pixel, so that vector graphics are smaller in size that bitmapped images. Vector files are spatial resolution-independent in that they can be resized easily without degradation in image quality. This file type is better suited to working with objects, such as a simple line or a shape, and for functions such as reformatting an object, like changing a red line to a blue line. There are several vector based graphic formats used on the Internet, but many are patented, and displayed with a "plug-in" or a helper application to assist the web browser, and thus have not been as popular as bitmapped formats for HTML-based projects [4-5]. This will likely change in the future, as increasing broadband access allows the rich interactivity of these files formats to be utilized in DTF work.
Scalable Vector Graphics (SVG, .svg)
One of the most popular vector-based graphic file formats for Internet DTF use is the Scalable Vector Graphics (SVG) format. This format is specifically based on XML in an effort to create a smaller, scalable image file format for Internet use. This file format has the advantages of vector graphics in that it is generally smaller than similar bitmapped files and scalable without loss of resolution. As the information for the image is stored in text data, it can be searched and edited, such as changing only the color of text in a diagram, which would be more difficult in a bitmapped image [4-5]. This information stored within the image allows for greater interactivity, with labels, legends, etc. on images. This is referred to as "metadata", or data about data, and allows for information such as a patient name and/or medical record number, as well as information about how the image should be displayed in a Web browser.
Bitmapped graphics, also called raster images, utilize pixels (picture elements) to define an image as a matrix made up of rows and columns of pixels, or dots. These file formats are preferable in DTF work since they are specifically designed to handle CT, MRI, and photographic images. The disadvantage of using pixel-based images is that they are resolution-dependent, such that the overall image quality is degraded if the resolution is altered.
Graphic Interchange Format (GIF, .gif)
The Graphic Interchange Format (GIF, pronounced "jiff", although "giff" is more often used) image format is the oldest, and most widely supported and common Web-based graphic file format, first created in 1987 [1, 4]. The primary strength of GIF is its lossless compression algorithm that provides up to 4:1 lossless compression of images. The GIF format also offers an interlace option that allows a rough, or preliminary version of an image to be displayed before the entire image is transmitted in a web browser [4, 6].
One weakness is GIFs limitation to a maximum of 256 colors, or shades of gray. GIF is a bitmapped image format with pixel values stored in image files. To explain this storage system, it is important to remember that the basic computer terms for data storage are the "bit" (Binary Digit, "1" or "0") and the "byte" (Binary Term), a unit of storage that on most modern computers consists of 8 bits.
Therefore, a 1-bit image would define each pixel as either black or white, with only two available shades (21 = 2), and an 8-bit image, can represent 256 different color values (28 = 256). GIF uses the classic RGB method for defining pixel values, where red, green, and blue values are each defined from 0 to 255, such that "0,0,0" codes for black and "255,255,255" codes for white.
The full RGB palate of colors contains millions different colors. To limit the size of files, however, GIF only allows 256 different colors, and keeps these 256 colors values indexed in a look-up table.
Because of this limitation, complex images may lose some detail when translated into GIF. The practical implication of this weakness of GIF is the inability to display an image using its full color range [4, 6]. GIF has been supported in most web browsers since 1990, and the two most widely used, Netscape Navigator (Netscape Communications, Mountain View, CA), since version 2.0 and Microsoft Internet Explorer (IE, Microsoft, Redmond, WA), since version 3.0 [6].
Joint Photographic Experts Group (JPEG, .jpg or .jpeg)
Joint Photographic Experts Group (JPEG, pronounced "jay-peg") was created in the early 1990s, and is not actually a file format, but the name of the compression algorithm. The JPEG image file format is referred to as JPEG File Interchange Format (JFIF, pronounced "jay-fiff"), and was specifically created for the storage and transfer of photographic images [6]. JPEG also uses bitmapped information stored in pixel files, but does not use indexed color. The 24-bit color of JPEG files displays each pixel on the screen with 24-bits of data encoding, allowing more color and contrast resolution than GIF (16.7 million colors, as opposed to 256 with GIF) [6-9].
The strength of the JPEG format is seen in its ability to significantly compress larger image files. Image compression up to 1/20 of the original file size can be achieved without significant loss of resolution [6, 8]. Such compressed images can be moved in the electronic environment with considerably more speed and would require significantly less storage space. JPEG also has a progressive display, although it is only available with newer web browsers, implemented in Navigator v. 3.0 and IE v. 4.0 [6, 7].
A weakness of JPEG in DTF work is the lossy compression technique, which may translate into image degradation and a possibility of distortion created by the technique. This can affect images, which are comprised of only a few colors, or those with large areas of the same color, and a Gibbs phenomenon that can be seen with high-resolution images [6, 10].
Tagged Image File Format (TIFF, .tif or .tiff)
The Tagged Image File Format (TIFF, pronounced "tiff") file format was developed in 1986, primarily by developers of input and output devices (Microsoft and Aldus - now Adobe), such as printers, monitors and scanners, so it is specifically designed to be flexible between different image processing devices [11].
The "tag" portion of the TIFF name refers to the complicated file structure, with groups of data called tags, which convey the image information to the displaying program. The actual specifications define over 70 different tag types, allowing for great flexibility between viewers. Although many programs simplify this by only implementing certain tags, skipping some tags could affect the image quality.
The strength of TIFF is that it can support any range of image resolution, size, and color depth and different compression techniques. The primary weakness of TIFF is the large file size that limits use in the Internet educational environment and slows overall performance. The newest specification incorporates the JPEG lossy method of compression, allowing for decreased file size [11]. Because of this large image file size, TIFF images are rarely utilized for Web DTF projects.
Portable Networks Graphics (PNG, .png)
The Portable Networks Graphics (PNG, pronounced "ping") file format was created in 1995, to define a flexible file format that allows lossless compression and gamma correction for cross-platform brightness consistency. A large push for this creation followed the move by certain companies to try to license the compression algorithm used by GIF images, and then collect royalties from software companies using the algorithm. Graphic artists and Web developers were determined to create an alternative to GIF that used a non-proprietary compression scheme. The data organized in a PNG file can contain keywords and strings of information, sometimes referred to as metadata, similar to SVG files, above, which is interpreted by certain decoder programs while being ignored by others [10, 12, 13]. The PNG chunk architecture supports three pixel types: palette-mapped, grayscale, and true-color (RGB) [10, 14]. PNG supports true color with true 48-bit images and 16-bit grayscale images [8, 10].
The PNG format utilizes a 2-dimensional interlacing scheme, which progressively displays the image even faster than a GIF image file [10, 14]. Gamma correction (this refers to the brightness reproduction curve of an image) allows the values displayed on any platform to be the same as the original, a powerful tool for use in viewing radiologic images [8, 10, and 14]. GIF and JPEG files have no inherent gamma correction [8, 10, and 15]. The two main browsers have had limited PNG support with early versions, and native support since versions 4.04 for Navigator, and 4.0 for IE [16].
Digital Imaging and Communications in Medicine (DICOM, .dcm)
The Digital Imaging and Communications in Medicine (DICOM, pronounced "dye-com") standard is the most comprehensive version of an imaging communications standard developed by the American College of Radiology (ACR) in conjunction with the National Electrical Manufacturers Association (NEMA). The purpose of DICOM is to provide platform-independent methods for connecting types of digital medical imaging devices. The standards are generated for PACS, with incorporation of detailed technical and demographic data, as well as methods for generating work lists and interacting with hospital information systems [17].
The strength of DICOM is its use in the PACS environment. These specifications allow for detailed descriptions of bitmapped digital images, including 12-bit grayscale as well as color, and annotations. In addition to complete image specifications, DICOM also prescribes multi-layer communications and network protocols that are specific for use in PACS [17]. The additional data in the DICOM specification that relates to communication protocols make this file format unwieldy for use in multimedia teaching tools outside of PACS. Although it is now widely accepted in the radiology community, the DICOM standard is relatively obscure in the non-medical digital imaging community.
The advances in hardware, software, and network speeds will lead to exponential growth of the number of DTF projects. The up-front organization of these projects is crucial for successful development and use. The many choices for image file formats, with various spatial and contrast resolution, and compression techniques must be considered for the project.
The four vital steps (acquisition, storage, manipulation, and presentation) in digital imaging must be considered prior to organizing such a project. The original acquisition of the image can be from the analog world, such as scanning a slide, or taking a picture of a hard copy film with a digital camera; or from the digital world, such as directly exporting an image from PACS. It is recommended that the image be stored prior to manipulation, if storage is available. Long time storage devices, external hard drives, CD-ROM, ZIP discs, or DVD, are relatively inexpensive, and is well worth the cost, especially when system failures can destroy DTF projects. The storage organization in the digital world is much more flexible than the original ACR code, usually following an anatomic/pathologic system, which is up to the user. More information on this can be found in other sections of the ASNR ELC syllabus.
The manipulation of the image should follow the storage, as it is often easier to re-edit an image, than it is to find out where you got the original from and then re-digitize an image. There are many image manipulation software tools available. The most popular is Adobe Photoshop, although the cost may be a limiting factor to some projects. The final step, the presentation can also follow multiple routes with a wide range of interactivity, from basic freeware image viewing programs, to complex Extensible Markup Language (XML) databases, utilizing cascading style sheets (CSS), with PHP: Hypertext Preprocessor (PHP). More information about this step can also be found on the ASNR ELC syllabus.
The limitation of contrast resolution with GIF images hinders their use in DTFs. While the TIFF format is versatile, the tagged architecture makes the image files too large for effective multimedia teaching projects. While JPEG files can be used effectively in Internet projects, the continued loss of data with each compression/decompression routine makes this format type less desirable for projects where images might be edited for use. The lossy compression technique is also not a good choice for the storage of raw data, for possible future use. Although DICOM is an effective media for the PACS environment, the multilayered communications protocols make this file format too bulky for DTFs. The PNG file format has superior contrast and spatial resolution, with gamma correction for cross-platform use, superior lossless compression algorithm (even compared to GIF), and faster display time in web browsers, making it an effective choice for some DTF work. The ability to store metadata within an image file is also an attractive feature work DTF projects.
Regardless of the image file format type chosen for a DTF, the importance of planning for the steps in digital image use (acquisition, storage and presentation) as well as image manipulation methods cannot be underestimated. This is just as important as the data field organization and relationships of fields in the DTF database.
 |
 |
| A. GIF Image | B. PNG Image |
 |
 |
| C. JPEG high res Image | D. JPEG low res Image |
 |
 |
| E. SVG Image | F. TIFF Image |
Comparison of common image file formats: All images were obtained from an original TIFF file that was exported from a Windows-based PACS. This original image was then opened in Adobe Photoshop v 6.0.1, as a 512 (width) by 724 (height) pixels, for a total of 362K pixel dimensions, with a document size of 18.06 cm by 25.54 cm, at 72 dpi. This image was then cropped to 511 by 566 pixels, and then was converted to 300 dpi and 600 by 665 pixels (2 by 2.17 cm). The image was then saved as a TIFF file without LZW compression (IBM PC byte order), and then saved as the other file formats. This was saved without an interlacing row order on the GIF and PNG files, no filtering on the PNG file, maximum quality (minimum compression) on the JPEG high resolution file (JPEG high), minimum quality (maximum compression) on the JPEG low resolution file (JPEG low), using Adobe Photoshop v 6.0.1. The cropped TIFF file was then opened in Adobe Illustrator v 9.0.1, and exported as an SVG file without compression.
This resulted in these file sizes:
Or, to list the files in order of increasing size:
| Image File Format | SVG (.svg) | TIFF (.tif) | JPEG high (.jpg) | JPEG low (.jpg) | GIF (.gif) | PNG (.png) |
| File Size | 63 KB | 395 KB | 116 KB | 16 KB | 195 KB | 139 KB |
So the JPEG compression algorithm is able to achieve an impressively small file size, while maintaining adequate image spatial resolution and the PNG file is actually smaller than the corresponding GIF, with higher contrast resolution. The SVG file is relatively small compared to the others, and is spatial resolution independent (it can be resized without loss of spatial resolution, which the other file formats are sensitive to when re-sized).
One interesting aspect to these many image file formats that relates to academic use is their use in Microsoft PowerPoint, which is used for 99% of lectures given digitally at major conferences. For example, if a user were to take a color image such as this one:
in Adobe Photoshop and save it as different image file formats, the file sizes would change as expected from the above discussion:
| Image File Format | Photoshop (.psd) | TIFF (.tif) | JPEG high (.jpg) | JPEG low (.jpg) | GIF (.gif) | PNG (.png) |
| File Size | 1.39 MB | 702 KB | 240 KB | 18.7 KB | 178 KB | 512 KB |
The original Photoshop file was 1.39 MB in size and the other sizes after saving in the different image file formats are listed above. The different JPEG options demonstrate the differences in files size with saving an image at high resolution (high) or low resolution (low).
If a user were to open PowerPoint and create a new slideshow file, and then save the file, without adding any text or images, the file size is approximately 8 KB, depending on the size of the hard drive on the system, and partitioning settings. The addition of a basic design template to a single slide will increase that file size from 8 KB to 52 KB.
If each of these image file formats are then inserted into an 8 KB PowerPoint slide show, the file sizes are changed to a certain degree because of the program, and if the PowerPoint file is then saved without any other formatting changes or additional text or images, the resulting PowerPoint file size will reflect this difference:
| Image File Format | Photoshop (.psd) | TIFF (.tif) | JPEG high (.jpg) | JPEG low (.jpg) | GIF (.gif) | PNG (.png) |
| File Size | 1.58 MB | 582 KB | 279 KB | 55 KB | 192 KB | 553 KB |
The Photoshop file was copied and then pasted into PowerPoint, as PowerPoint will not allow a .psd file to be inserted from file. So, you might expect that the resulting file size would also increase by 8 KB, being the size of the image “inserted from file” within PowerPoint, and the size of an unformatted PowerPoint file. As you can see by comparing the charts, however, there is an unexpected change in the resulting file size:
| Image File Format | Photoshop (.psd) | TIFF (.tif) | JPEG high (.jpg) | JPEG low (.jpg) | GIF (.gif) | PNG (.png) |
| File Size Change | + 19 KB | - 120 KB | +39 KB | +36 KB | +14 KB | +41 KB |
The overall differences in size are due partially to the type of image file format, but there is one surprising change in this chart. The TIFF image file format in the PowerPoint slide show file actually decreases in size in comparison with the original image file format itself. The feature of the TIFF file format described above explains this. The PowerPoint program attempts to incorporate files inserted such that they become similar to a Microsoft embedded object, but the metadata within the TIFF image file format is not needed for basic display purposes, so PowerPoint is actually stripping this data away for the purposes of the use of the image in PowerPoint.
The user should always be aware that the use of images in PowerPoint is very different from the use of audio or video files inserted into a PowerPoint file, in that PowerPoint is essentially creating a pointer that leads the program to the audio or video file, and not inserting it into the PowerPoint file. The user must therefore be careful to always move the audio or video file with the PowerPoint talk, and to keep them in the same relationship in the hierarchal file folder system used. One way to prevent problems is to create a folder for the talk including the PowerPoint file and the associated audio and video files, and to always be careful to move all files in the folder together, by moving the folder itself, instead of moving only the PowerPoint talk file independent of the supporting files.
Another important factor of image file formats, in addition to size, is their inherent attributes:
Image File Format |
GIF |
JPEG |
TIFF |
PNG |
Maximum Color Depth |
8-bit |
24-bit |
48-bit |
48-bit |
Compression Technique |
Lossless |
Lossy |
Lossless(Lossy) |
Lossless |
Browser Interlacing |
Yes (1-D) |
No |
No |
Yes (2-D) |
Gamma Correction |
No |
No |
Yes |
Yes |
Comparison of GIF, JPEG, TIFF and PNG attributes, adapted from [8, 10, 15, 16]. See discussion above.
ASNR 2001 ELC Editorial board questions and responses:
Q. Since .tif involves large file size what would we lose by not even having it - is it that much better than .jpeg for or .png for transportability? I like the Kodak Imaging for Windows on my computer for multifile .tif - for black and white imaging. That is really the only application for it that I use. I haven't checked, but for that purpose is it superior and is the file size adequate compared to the other file types? This isn't Radiology necessarily but it is a useful application for anyone. Please comment.
A. TIFF was specifically made by the people making the input and output devices, so they preferentially used it for their products, making it more popular, but, no, you probably aren't losing much using JPEG. It is higher resolution, 48-bit, and it uses a lossless compression algorithm, so you aren't loosing anything opening and closing the files. And sometimes a TIFF image might print out better on older printers when moving a file between a flat-bed scanner and computer then to a printer, but there usually isn't much of a problem with newer equipment. If there is an image that you are going to open, edit, save, repeat, though JPEG isn't a very good choice because you are losing a little bit of data each time you do that. But it is OK to work with an image as a TIFF or PNG, and then save it as a Web optimized JPEG for final versions of multimedia projects. The web optimized version involves limiting the number of colors in an image for display in a web browser; actually the web browsers prefer to display images in a "web safe" palette of about 254 colors.
Yes, the Windows imaging default imaging program on the Windows platform is good for basic things and even for transferring images from PACS to the o clipboard to another program, but Photoshop is really the gold standard for imaging editing, and I like Debabelizer for image translation.
The use in web browsers limits TIFF in DTF work, because the files are so big and because there is no progressive display like GIF and JPEG and PNG, and those TIFF images have to be displayed with the help of a helper application through IE or Navigator.
This answer is a little long-winded, but I hope it answers your questions.
Q. Can you refresh us on Gibb's phenomenon?
A. Gibb's phenomenon exists in the digital imaging and photography world, but for us, as neuroradiologists, it is the bright abnormal signal that you might see on sagittal T2 weighted images through the spine, that is not there on the axials or T1 sagittal views. This is caused by the drastic change in intensity between the cord and the surrounding CSF.
ASNR 2002 ELC Editorial board questions and responses:
Q. Can PowerPoint accept all image formats?
A. Essentially Yes, Microsoft PowerPoint will take just about any image file format in the new versions. Microsoft Office 2002 (XP) specifically lists: .emf, .wmf, .jpg, .jpeg, .jfif, .jpe, .png, .bmp, .dib, .rle, .bmz, .gif, .gfa, .emz, .wmz, .pcz, .pcd, .pcx, .cdr, .cgm, .eps, .fpx, .mix, .pct, .pict, .tif, .tiff, and .wpg. under types of image file formats when inserting an image from file.
Q. How do you organize your files?
A. We have saved images from our PACS as TIFF images, since that was the easiest methods to export them, from our Windows based system. We developed an anatomic/pathologic landscape type architecture for case storage, with spine and head and neck cases starting with anatomy, and brain cases starting with pathology, since anatomy doesn't always have the same implications to diagnoses in the brain as it does in the head and neck and spine.
Q. Do you know of a program that can organize, search, and perhaps display thumbnails for images?
A. I have tried a few programs, like the digital assent management tool, Cumulus (from Canto: http://www.canto.com ), but have now upgraded to Windows XP, which, like Windows 2000, gives you the option of viewing the contents of a folder as thumbnails in the Windows Explorer windows (along with Tiles, Icons, List, and Details). So if your pathologic/anatomic landscape architecture is set up so that a particular case is in a folder organized in whatever scheme you like, with the relevant information, you should be able to find interesting cases by digging down through the folders. I.e. to find an acoustic schwannoma, I would click on "DTF/Head and Neck/CPA/Benign tumor", and would then be in the folder with all my acoustic schwannoma cases.
Alternatively, to search the folder, it is easy to "map" a particular drive if it is on a network, and then search the DTF for the diagnosis. I.e. to find an acoustic schwannoma case, I know that I have each case image files in a folder labeled acoustic schwannoma, with other relevant info, name, medical record number, and basic information, such as classic, atypical, cystic, etc. I can just search in Windows Explorer for "Acoustic Schwannoma", and the computer will find all the folders with that name, and I can look at the list to open the classic, atypical, etc. case I am interested in.
Image File Format Acronyms:
Image File Format Glossary:
Image File Format Suffix List:
Copyright © 2004 American Society of Neuroradiology, www.asnr.org