Method for obtaining high-resolution performance from a single-chip color image sensor
View Patent ↗A method is provided for obtaining high-resolution performance from a standard image sensor, such as a single-chip image sensor. Since single-chip color image sensing is often the least expensive way to obtain color images, the invention helps to lower the cost of high-resolution machine vision performance. The color image produced by the single-chip image sensor is processed in two ways at run-time. In one stage of run-time, processing is performed on the original color image provided by the image sensor, including normal low-resolution image processing and low-resolution machine vision analysis. In another stage of run-time, a high-resolution search model is determined to search the color image using knowledge of the geometric properties of the arrangement of the pixels of the low-resolution image sensor, and the sub-pixel phase of the high-resolution search model. The invention can as much as double the accuracy of machine vision analysis using a standard color camera.
1. A method for high-resolution (high-res) location of an object in an image using a single-chip color image sensor that is sensitive to N bands of electromagnetic energy, the sensor having a sensor geometry, comprising:
providing a high-res color model of an object;
acquiring a color image of an object with the color image sensor;
finding an alignment location of the high-res color model in the color image so as to provide a single low-res location having a phase, including providing N low-res single-band locations, one low-res single-band location for each of the N bands, and combining the N low-res single-band locations into a single low-res location;
using the phase of the single low-res location, and using the sensor geometry, so as to provide a true-phase high-res color model having correct phase; and
finding a high-res location of the true-phase high-res color model in the acquired color image.
2. The method of claim 1 , wherein combining the N low-res single-band locations includes:
combining the N low-res single-band locations into a single low-res location using a weighted sum, each location being associated with a weighting factor.
3. The method of claim 2 , wherein the weighting factor includes a confidence factor.
4. The method of claim 1 , wherein the sensor geometry is based on the Bayer pattern.
5. The method of claim 1 , wherein providing a high-res color model of an object includes:
creating a high-res synthetic model of the object.
6. The method of claim 1 , wherein providing a high-res color model of an object includes:
acquiring a high-res image of the object using a high-res camera.
7. The method of claim 1 , wherein providing a high-res color model of an object includes:
acquiring a low-res image at each phase inherent in the sensor geometry.
8. The method of claim 1 , where N is 3, and the bands represent red, green, and blue.
9. The method of claim 1 , wherein using the phase of the single low-res location, and using the sensor geometry, so as to provide a true-phase high-res color model having correct phase includes:
selecting one of N candidate high-res color models by using the phase of the single low-res location.
10. The method of claim 1 , wherein using the phase of the single low-res location, and using the sensor geometry, so as to provide a true-phase high-res color model having correct phase includes:
synthesizing a synthetic model.