TY - JOUR
T1 - Color images in telepathology
T2 - How many colors do we need?
AU - Doolittle, Marc H.
AU - Doolittle, Keith W.
AU - Winkelman, Zev
AU - Weinberg, David S.
PY - 1997
Y1 - 1997
N2 - It is generally assumed that for telepathology, accurate depiction of microscopic images requires the use of 'true color' (ie, 24 bits, eight bits each for red, green, and blue) in the digitized image used for transmission. If such a 24-bit color image file, which provides a palette of 16.7 million colors, could be reduced in size by decreasing the possible numbers of colors displayed in the image to g hits (palette of 256 colors), the image files would require less storage space, could be transmitted more rapidly, and would require less telecommunications bandwidth. However, such color reduction must not result in detectable image degradation, especially if the images are to be used for diagnosis. Therefore, we performed a carefully controlled study to determine whether pathologists could detect differences in the quality of microscopic images that were reduced from 24 to 8 bits of color. Thirty, pathologists were each asked to view a set of 30 image pairs displayed on a computer monitor. Each image pair consisted of the original 24-bit color version and an 8-bit color reduced version, derived using an adaptive color reduction algorithm with diffusion dithering. Observers were asked whether they could detect any difference in quality between the image pairs. Then, regardless of their answer, they were asked to choose the better quality image of the pair. Overall, there was not a statistically significant ability to consciously detect differences between the image pairs (P < .750). However, when forced to choose, there was a significant preference for the 8-bit images as being of 'better quality' (P < .005). We conclude that telepathology applications may be able to take advantage of adaptive color reduction algorithms to reduce image file size without sacrificing image quality. Additional studies must be performed to determine the minimal image requirements for accurate diagnosis by telepathology.
AB - It is generally assumed that for telepathology, accurate depiction of microscopic images requires the use of 'true color' (ie, 24 bits, eight bits each for red, green, and blue) in the digitized image used for transmission. If such a 24-bit color image file, which provides a palette of 16.7 million colors, could be reduced in size by decreasing the possible numbers of colors displayed in the image to g hits (palette of 256 colors), the image files would require less storage space, could be transmitted more rapidly, and would require less telecommunications bandwidth. However, such color reduction must not result in detectable image degradation, especially if the images are to be used for diagnosis. Therefore, we performed a carefully controlled study to determine whether pathologists could detect differences in the quality of microscopic images that were reduced from 24 to 8 bits of color. Thirty, pathologists were each asked to view a set of 30 image pairs displayed on a computer monitor. Each image pair consisted of the original 24-bit color version and an 8-bit color reduced version, derived using an adaptive color reduction algorithm with diffusion dithering. Observers were asked whether they could detect any difference in quality between the image pairs. Then, regardless of their answer, they were asked to choose the better quality image of the pair. Overall, there was not a statistically significant ability to consciously detect differences between the image pairs (P < .750). However, when forced to choose, there was a significant preference for the 8-bit images as being of 'better quality' (P < .005). We conclude that telepathology applications may be able to take advantage of adaptive color reduction algorithms to reduce image file size without sacrificing image quality. Additional studies must be performed to determine the minimal image requirements for accurate diagnosis by telepathology.
KW - digital imaging
KW - telemedicine
KW - telepathology
UR - http://www.scopus.com/inward/record.url?scp=0031038592&partnerID=8YFLogxK
U2 - 10.1016/S0046-8177(97)90276-8
DO - 10.1016/S0046-8177(97)90276-8
M3 - Article
C2 - 9013829
AN - SCOPUS:0031038592
SN - 0046-8177
VL - 28
SP - 36
EP - 41
JO - Human pathology
JF - Human pathology
IS - 1
ER -