Research And Design Of Nonlinear Scaling Circuit For Video Images Base On Interest-region Detection

Nonlinear scaling circuit for video images base on interest-region detection is oneof key module of the multi-standard video post processing chip. The subsystem ismainly responsible for converting the source video signal to the format configured byuser, its main function includes image scaling, skin tone adjusting, output timinggenerating and so on.While format conversion, it maybe emerge that incompatibility problem betweenimages and display format on their wide-high scale. Distortion occurs when image isresized by linear scaling method. According to the existing scaling methods, an imageis divided into several blocks to ensure that the central region of the image is notdistorted. As a result, main spot near the border may be distorted because of the greatchange in interpolation points density. Based on human visual habits, a novel methodof nonlinear scaling is proposed with skin tone and Harris corner detection. Variablelocation of the nonlinear part was used to strive for the main spot was not distorted.The propose scaling method are evaluated by introducing roundness of interestingregional characteristic circle, analysis shows that the main spots of the results weredistorted less slightly than other method.In proposed nonlinear scaling circuit, there are3kinds of scaling modes:proposed nonlinear scaling, linear scaling and letter-box, user can configure it throughexternal interface. Only few line buffers are used instead of frame buffer to reduce thearea of circuit. However, data of whole frame are needed for interest-region detection.Detection result of current frame is used to nonlinear scaling of the next frame forslightly difference between two adjacent frames. By analyzing some interpolationarithmetic, different interpolation algorithms are used in rows and columns. Thisdissertation adopts cubic method to rows interpolation for performance and linearmethod to columns interpolation for fewer line buffers.Finally, results of function simulation and verification show the proposed videoimage process subsystem has achieved the design target basically.