Demosaicking Based on Direction Categorization in Wireless IP Camera

A new demosaicking algorithm based on direction categorization for wireless IP camera is proposed. By calculating the differences of neighboring pixels and the difference of average value of horizontal and vertical green pixels, the centered pixel is classified as horizontals verticals smooth and non-direction. After the directionality of the centered pixel is determined, the gradients of adjacent pixels along this direction are employed to gain the interrelated pixels and the secondary interrelated pixels which are adopted to calculate Kr/Kb of the centered pixel and then interpolate the missing elements. Compared with other CFA demosaicking methods, the proposed method can greatly reduce the artifacts causing by interpolating across edges. Not only can this algorithm obtain high PSNR, but also can it maintain low hardware cost according to the experimental results.