Edge detection in color images is generally based on output fusion methods, gradient methods, and vector methods. In output fusion method, the edge detection is carried out three times, once each for red, green, and blue (or any other choice of color space), and then the output is fused to form a single edge map. Multi-dimensional gradient methods are used to reduce the workload by combining the three gradients into one and later detecting only the edges. In vector-based methods, the whole image is considered as a vector field and various statistical operations are used for edge detection.
The authors of this Applied Optics article propose a new method of edge detection in color images, based on the novel use of a digital light projector incorporating Digital Micromirror Device (DMD) technology. They use a property of the DMD to simultaneously generate a positive and negative replica of the same image. For a color image, the negative copy corresponds to its complimentary color image. These two complimentary images are then overlaid after a slight displacement in a given direction resulting in a profound edge enhancement that is associated with partial derivatives along the displacement.
DMD is an array of thousands of very small switchable mirrors, whose speed, contrast ratio, and wide spectral response makes it an excellent spatial light modulator. The novel applications of DMD’s are beyond the common projection displays (video projectors) and are not limited to 3D display, lithography, holographic data storage, microscopy, medical imaging, and computer vision.
When a DMD is used in a light projector for digital image display, the micromirrors can have two positions, allowing the light reflected by the mirrors to reach the display screen or to be blocked inside the projector. The micromirrors are moved at a high frequency producing binary single color images in succession at the output screen that are then perceived as half-toned color images.
In their adoption of a digital light projector, the authors make use of the otherwise discarded light by the DMD into what they call “a complimentary color image”. The two states of the micromirrors from the DMD are used to produce a “positive” and a complimentary “negative” replica of a color image by illuminating from two opposing sides. The resulting two images are made to overlap at the output screen by shifting slightly one of the images in a lateral direction, thereby, creating an incoherent superposition of the two images with a partial derivative in the direction of the lateral shift. The end result is edge enhancement of the color image in the direction of selected lateral shift.
The proposed method of edge detection in color images is based on superposition of images directly without prior decomposition of images into RGB space and then using gradients or digital operations. The absence of any numerical processing to achieve edge enhancement in this technique has the potential to process large color images in real time edge detection applications.
You must log in to add comments.