The watershed algorithm is a powerful morphological tool for image segmentation

but the traditional watershed algorithm causes serious over-segmentation if the actual image is affected by specularities and shadows. To solve these problems

a new image segmentation

based on reconstruction labeling watershed algorithm in color space

is proposed. First

the RGB color image is converted to a new color space image. The directions that are not influenced by specularities and shadows are extracted to calculate their gradients. Then

object regions are extracted by using the morphological opening and closing to compose a binary marked image

and the gradient image is introduced to substitute the marked image. Finally

the watershed transformation is employed to the modified gradient image. The new algorithm cannot only overcome over-segmentation

which is produced by texture details and noise

but also can suppress over-segmentation caused by specularities and shadows. Moreover

the segmentation algorithm is executed on a primitive gradient image instead of filtering and simplified image

so the object's edge information is retained greatly. Theory analysis and experimental results have shown the segmentation algorithm is effective.