Efficient Color Histogram Relationship Matching Approach Based on Absolute Heavily Dependent Spatial Patterns

The successful application of any vision system is reliant on the right choice of low-level features. For model-based visual context, we must compare sub-regions of an image. Thus, the features selected will represent a collection of pixels. The following is not exhaustive, but is a start: Color: With proper extraction, color is one of the most reliable properties. If color is isolated from intensity, it tends to remain constant under a wide variety of natural lighting conditions. In some applications, either mean or mean and variance may suffice. For more careful matching a color histogram may be necessary. Intensity: From image to image the intensity of the same object can vary greatly.