Performance Evaluation of Line-Sensor 3CCD Camera and Comparison of Specialist's Determination and Numerical Approach

Some cameras, especially in the case of assembling line-sensor 3CCD cameras, may generate non-suitable images which include phenomena called "pixel shifts." This paper describes such a phenomenon, its numerical analysis and comparison between professional determination and machine detection. If there is such a pixel shift in the image, a user of camera may be suffering from a lack of precision of the image to be measured or utilized for image understanding. In a conventional approach of manufacturing line-sensor 3CCD cameras, some specialists have determined whether phenomena of "pixel shifts" occurs or not in the image. They have reported performance evaluation for the relevant camera which monitors the image before shipping camera. We had already proposed a numerical method to detect occurrence of pixel shift in the image with Lagrange interpolation function, calculated the values of the level for occurrence of pixel shift quantitatively. In the previously comparison, our calculated results had been perfectly consistent with determination by some specialist. And moreover, we have applied our approach into another kind sets of 20 samples of camera-monitored image and compared determination by another specialist with our corresponding results. By means of Kansei Evaluation of human sensitivity, we have discriminated difference of specialist's determination and our numerical approach.