Quantitative evaluation of overall electronic display quality

Abstract

T HE USE OF ELECTRONIC display devicesfor primary diagnostic interpretation of digital images in Radiology is becoming commonplace. For all equipment involved in image acquisition or display, it is typical practice in radiology that quantitative measurements of physical parameters affecting image quality be made for the purposes of new equipment selection, acceptance testing, and quality control (QC). Electronic displays are problematic in that routine measurement of display parameters other than luminance are difficult and expensive to obtain, especially in the field. To address this difficulty, we are investigating a method for quantitatively evaluating the overall quality of electronic displays. This technique is very similar to traditional contrast-detail (CD) methods'? in that many factors affecting image quality (including contrast sensitivity, noise and modulation transfer) are included in an overall quality measure. A technique acceptable for the purposes of new equipment selection, acceptance testing, and QC would allow the quantitative evaluation of a device in 30 minutes or less with a precision of about 10%. Such a technique would also allow quantitative evaluation of the effects of display set-up and environment, such as maximum display luminance and ambient room lighting.The purpose of the current work is to make an initial pilot investigation of the feasibility of using contrast-detail techniques to provide overall quantitative quality evaluations of electronic displays. Our data may also allow us to make a preliminary evaluation of the importance of maximum display luminance. METHODSContrast-detail methods 1-3 involve the presentation of simple targets on a uniform background to a group of observers. The targets have varying contrast and size. For each target size, the observer indicates the contrast threshold at which the target is just visible, and in most cases, interpolation between actual image contrast is encouraged. We incorporated the idea of requiring the observer to indicate the area of the test image in which the target is seen as proof that the target is actually visualized.' Six sets of 8-bit per pixel test images were created. Each set consisted of eight images, each image corresponding to one of eight possible square target sizes. The target sizes were 1,2,3,4, 7, II, 17, and 27 pixels. In each test image, eight rows of four test areas were present. Each row corresponded to one of eight possible target contrasts. The target contrasts were I, 2, 3, 4, 7, II, 17, and 27, denoted as the pixel value difference between the target and the background. For object size of I pixel, larger contrasts were used. All targets were darker than the background to enhance sensitivity to cathode ray tube (CRT) glare. Each test area presents the target in one of four randomly selected quadrants. The test areas are defined by low contrast borders. A test image set is schematically illustrated in Fig I. The test image sets were created using a common PC paint application (PaintShop Pro. Version 4, lASC I...

Effects of monitor luminance change on observer detection performance

“…We calculated the rate of correct detections of observers in the above-mentioned 4-AFC task, and found the display conditions in which this correctly decision rate showed statistically significant inferiority compared with the brightest condition (the 11th CRT display condition). Here, in the calculation of correct detection rate, a score of 1 was assigned for each correct response, 0 was assigned for each incorrect response, and 0.25 was assigned for each 'no target seen' response [10]. For our experimental design, a statistical analysis of the differences among the 11 display conditions was needed.…”

“…To evaluate the effect of changes in CRT monitor luminance on the detection performance, the image-reading [9], and the design for test images was analogous to the contrast-detail method proposed by Hangiandreou et al [10,11]. The luminance changes in CRT monitor displays were the same as in our previous study, and were the ones in which the grid voltage was decreased [8].…”

Optimization of a contrast-detail-based method for electronic image display quality evaluation

“…8 Three test sets of 8-bit-per-pixel test images were created using a common PC paint application (PaintShop Pro, Version 4; JASC Inc, Eden Prairie, MN). Each test set consisted of eight images, each image corresponding to one of eight possible square target sizes (1,2,3,4,7,11,17, and 27 pixel edge lengths). In each test image, eight rows of four test areas were present.…”

Quality Control of Display Devices for Medical Radiodiagnosis Employing the Standard TG18, GSDF and JND

“…If the noise test identifies some behavioral failure, the luminance and contrast response should be verified [4]. The test consists of observing a series of small frames containing small low-contrast structures in each quadrant of the image display device [7]. Thus, the existence of artifacts or defective pixels that could compromise the visualization of small structures with a low-level contrast [15,16] is verified.…”