New Morphological Algorithm by Using Conspicuousness Map for Color Image Processing

Abstract

Morphological image processing is well known as a powerful method in image processing and computer vision. This paper proposes a novel morphology algorithm in consideration of the visual characteristic of a color image. In the traditional morphological processing of the color images, the magnitude correlation of the pixel vector is determined beforehand and has been applied to any color images. Therefore, the effectiveness of the processing will vary depending on the contents of individual images. In this paper, a new concept of Conspicuousness Map of the notability of a pixel, suiting human feeling in the conspicuous grade between the pixels in the structure element of morphology, is proposed, aiming to apply to morphological processing. By the proposal method, the effect to reduce or remove a thin line noise or blob noise regardless of the image contents, which was impossible to acquire by the conventional morphological processing, is attained.