Proposal of a full color automatic natural image region segmentation method with no limitation to the number of regions and the complexity of the area
Abstract
In this paper, a full color natural image region segmentation method with no limitation to the number of regions and the complexity of the area was proposed, and the effectiveness was confirmed by some natural image pictures.
Automatic natural image region segmentation continues to be in high demand for a long time and this difficulty mainly depend on the large difference between the region division by the human perception and the result of the calculation theory approaches based on the color extraction. Also, the many region segmentation algorithms could work in the case of binary image. The proposed the mathematical procedure that (1) always area segmentation can be realized by only two times scan lines even an image took complex shapes, (2) it does not depend on the number of regions and (3) color and brightness evaluation functions are introduced. Experimental result shows that the proposed method can divide many complex shape areas which the number of regions is uncertain and our method can divide many objects with gradation patterns correctly in the natural full color image. Our method will reduce the complexity of the problem of recognition especially for the complexity growing by increasing the number of objects. It will play basic role for many image processing study fields.
Keywords
Download Options
Introduction
In this paper, a full color natural image region segmentation method with no limitation to the number of regions and the complexity of the area was proposed, and some natural image pictures confirmed the effectiveness. Automatic natural image region segmentation has been continued to be in high demand for a long time [1-21] in the large field of image processing. The difficulty is mainly divided to two points. (1) The number of regions is uncertain. (2) The shape of the regions is unstable. A set of algorithm not to be affected by the above two difficulties was proposed in this paper. Fig.1 shows the proposed algorithm process flow. It can realize automatic natural image region segmentation by two-scan line process (horizontal and vertical) and continuously called one simple replacement subroutine [22]. A typical proposed algorithm result was shown in Fig.2 and the area was segmented even if there is a gradation area of the color.
Conclusion
In this paper, a natural image area segmentation method and confirmed the performance with gray-scale, complex shape, colored gradation images was proposed. The mathematical procedure have three features that (1) always area segmentation can be realized even an image took complex shapes, (2) it does not depend on the number of regions and (3) the color and brightness evaluation function E is introduced. Our method's important feature is that it can be applied to the natural full color image automatic area segmentation. It will reduce the complexity of the problem of recognition especially for the complexity growing by increasing the number of objects and it will play basic role for all field of image processing study including fields.