|
On-Chip ProcessingTraditionally, the only thing a sensor did was measure incoming light. All other image processing operations were performed on a computer or some special purpose hardware. However, newer sensors such as Bigshot's Aptina MT9D111 [1] sensor perform demosaicing, color enhancement, image sharpening, and image compression before storing the image in memory. Here are some of the things that Bigshot's sensor can do: Auto-Exposure Bigshot's image sensor can adjust the amount of time that its pixels are exposed to light based on the brightness of the scene. If the scene is very dark, the sensor will expose the pixels for a longer time to receive more light. On the other hand, if the scene is too bright, it will reduce the exposure time so that the image will not get washed out (saturated) with too much light. This feature is called auto-exposure. First, the sensor captures a quick photo and calculates its average brightness to see how much light is in the scene. Then, the sensor chooses an exposure time for the actual photo so that it will not be too bright or too dark. If the scene is extremely dark, the sensor takes the extra step of triggering Bigshot's flash to brighten the scene. Image Compression A digital image, as described before, contains red, green and blue values at each pixel. Bigshot's image sensor has a grid of 1600x1200 pixels, and each color is represented using 8 bits (or 1 byte) of data. Therefore, the total memory required to store an image is (1600 x 1200 pixels) x (3 colors per pixel) x (1 byte per color) = 5.7 Megabytes, roughly 5,700,000 bytes! All cameras have a limited amount of memory to store images with. If we can reduce the amount of memory needed to store each image, then we can fit more images into the memory. This process of size reduction is called compression. To see compression in action, consider the sequence "AAAAAAAAAAAABBBBBCCCCCCAA," which is 25 characters long. Since it has 12 A's, 5 B's, 6 C's, and 2 A's, in that order, it can be written more compactly as "12A5B6C2A"; that is, with just 9 characters. This gives us a saving of 16 characters with no loss of information. One of the most popular compression techniques for images is JPEG (pronounced "jay peg") compression [2]. Bigshot's on-chip JPEG encoding circuit compresses each captured image to a size that is around 20 times smaller than the size of the original image without any visible loss of quality!
References
|