| Popis: |
The technique developed for experimental evaluation was an image catalog containing a standard set of 231 images having diverse scene content and quality. The image variants in the catalog were assigned indexes of image interpretability based on the measured performance of 154 interpreters. The catalog technique requires the interpreter to compare any new image with images in the catalog and to select the catalog image most similar in quality to the new image. The indexes associated with the selected catalog image are taken as measures of the interpretability of the new image. The catalog technique had a high degree of validity, particularly when estimates were provided by trained interpreters (average correlation coefficient for target area discrimination was .77, for target identification, .54). The technique was also effective with untrained subjects, although not as high (.70 and .51). The other techniques were also effective. The technique combining catalog judgments with weighted physical measures of image quality was best for target area discrimination (.86 vs .52 for target identification); the technique using physical measures alone was best for target identification (average validity coefficient of .60 vs .43 for target area discrimination). (Author) |
