A practical neural network model was designed to realize the color space conversion of digital photofinishing. The sampling, network structure and training process were introduced respectively. But in actual training, the networks fall into local minimum in all probability. To solve this problem, evolutionary programming (EP) algorithm was applied and the learning rate was adaptively adjusted. In the experiment, the performance of network was compared with pre-optimizing. Then the color space conversion was evaluated by the simulation error of samples from the point of color difference.

The equi-luminance of color stimulus in normal subjects is characterized by L-cone and M-cone activation in retina. For the protanopes and deuternopes, only the activations of one relevant remaining cone type should be considered. The equi-luminance turning curve was established for the recorded visual evoked potentials (VEPs) of the luminance changes of the red and green color stimulus, and the position of the equi-luminance was used to define the kind and degree of color vision deficiencies. In the test of 47 volunteers we got the VEP traces and the equi-luminance turning curves, which was in accordance with the judgment by the pseudoisochromatic plate used in clinic. The method fulfills the objective and quantitative requirements in color vision deficiencies test.

A novel method for spectral characterization of scanner was proposed in this paper, which combined the principal component analysis (PCA) and back propagation (BP) artificial neural network (ANN). The natural color system (NCS) color patches were adopted as the color targets. The accuracy of this method was evaluated by spectral root mean square (SRMS) error and the CIEDE2000 color difference specification. The experimental results showed that six principal components were appropriate and the spectral characterization accuracy was outstanding when a 3-20-6 BP net structure was used to estimate the scalars from the scanner red/green/blue (RGB) signals.

A novel virtual four-ocular stereo measurement system based on single high speed camera is proposed for measuring double beating wings of a high speed flapping insect. The principle of virtual monocular system consisting of a few planar mirrors and a single high speed camera is introduced. The stereo vision measurement principle based on optic triangulation is explained. The wing kinematics parameters are measured. Results show that this virtual stereo system not only decreases system cost extremely but also is effective to insect motion measurement.

In this letter, it is shown that there exists relationship between phase-based and area-based stereo in spite of their different motivations. A new cost function is defined based on this clue and an improved cost-minimization framework is presented. It is suitable for disparity estimation and occlusion detection in aerial scenes.