HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Full Text Full Text (PDF) Submit a response Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Slone, R. M. Articles by Hendrickson, D. D. Search for Related Content PubMed PubMed Citation Articles by Slone, R. M. Articles by Hendrickson, D. D.

Assessment of Visually Lossless Irreversible Image Compression: Comparison of Three Methods by Using an Image-Comparison Workstation¹

¹ From the Mallinckrodt Institute of Radiology, Washington University School of Medicine, Barnes-Jewish Hospital, Box 8131, 510 S Kingshighway Blvd, St Louis, MO 63110. From the 1998 RSNA scientific assembly. Received December 30, 1998; revision requested February 23, 1999; revision received July 16; accepted August 18. Address correspondence to R.M.S. (e-mail: sloner@mir.wustl.edu).

View larger version (94K): [in a new window]   Figure 1. Direct screen capture shows the ICW graphical user interface. A small replication of the entire image is shown in the lower left-hand corner, where the white box defines the portion of the image displayed in the top 80% (2,048 x 2,048 pixels) of the monitor. The set of resolution images for comparison is listed in the box on the right for test case 18-5. The "Prev" and "Next" buttons in the "Test Image" area change the test image being compared. The "Prev" and "Next" buttons in the "Control Image" area change the resolution reference image. This can also be changed by using the wheel on the mouse or the up and down arrows in the "Best Match" area. The "Flicker" area offers selection of "Auto" or "Manual" for automatic or manual control of flicker, respectively, and a slider to set the flicker rate in the automatic control mode. The "Zoom In" area offers a choice of x1, x2, or x4 magnification. The x2 magnification is displayed in this example. The "Mark as Best" button records the control image selected as the best match to the test case. A warning is displayed if the user tries to move to the next case without recording a choice.

View larger version (105K): [in a new window]   Figure 2. Compression artifacts. Compressed (JPEG, left column; WTCQ, right column) and noncompressed digital (middle column) posteroanterior radiographs of a selected region of interest show the effects of compression at ratios of 8:1 (top row), 16:1 (middle row), and 128:1 (bottom row). With both JPEG and WTCQ algorithms, the image compressed at 8:1 is indistinguishable from the original. At a compression ratio of 128:1, the manifestation of "tiling" or "blocking" artifacts on the JPEG image and blurring on the WTCQ image are readily apparent.

View larger version (100K): [in a new window]   Figure 3. Resolution metric. Composite image shows the spatial-resolution scale applied to a representative portion of a radiograph. Reduction in resolution progresses from grade 1 (top left and bottom right: unaltered original, 100%) to grade 15 (bottom row, second from right: 4%). The percentages are a measure of two-dimensional spatial resolution, which was determined on the basis of the square of the bandwidth reduction implemented with power-law filters and can be thought of as the percentage of pixels displayed.

View larger version (29K): [in a new window]   Figure 4. Bar graph shows the combined results for five observers in the two-alternative forced-choice experiments, with the percentage of test images judged to be superior to the original image. The horizontal line at 50% represents the expected result for a purely random selection. Error bars = 95% CIs, gray bars = WTCQ images, white bars = JPEG images.

View larger version (31K): [in a new window]   Figure 5. Bar graph shows the combined results for five observers in the original-revealed forced-choice experiments, with the percentage of test images classified as equivalent to an original image. Error bars = 95% CIs, gray bars = WTCQ images, white bars = JPEG images.

View larger version (29K): [in a new window]   Figure 6a. Resolution-metric matching with (a) JPEG images and (b) WTCQ images. Scatterplots show the relationship between compression ratio (CR) and percentage resolution for observer A. These data were obtained by matching the blur image set with 20 test images at each of five compression ratios (16:1, 21:1, 32:1, 64:1, and 128:1). The area of each dot is proportional to the number of superimposed data points. The straight lines are the regression lines. (a) The x intercept for JPEG images, determined by means of linear regression on the 100 data points, indicates that the estimated visually lossless threshold is 13:1. (b) The x-intercept for WTCQ images, determined by means of linear regression on the 100 data points, indicates that the visually lossless threshold is 11:1.

View larger version (28K): [in a new window]   Figure 6b. Resolution-metric matching with (a) JPEG images and (b) WTCQ images. Scatterplots show the relationship between compression ratio (CR) and percentage resolution for observer A. These data were obtained by matching the blur image set with 20 test images at each of five compression ratios (16:1, 21:1, 32:1, 64:1, and 128:1). The area of each dot is proportional to the number of superimposed data points. The straight lines are the regression lines. (a) The x intercept for JPEG images, determined by means of linear regression on the 100 data points, indicates that the estimated visually lossless threshold is 13:1. (b) The x-intercept for WTCQ images, determined by means of linear regression on the 100 data points, indicates that the visually lossless threshold is 11:1.