Apparatus and method of recognizing image feature pixel point

• US 8,180,170 B2
• Filed: 08/19/2008
• Issued: 05/15/2012
• Est. Priority Date: 12/31/2007
• Status: Expired due to Fees

• Current Assignee: Altek Corporation
• Original Assignee: Altek Corporation
• Inventors: Chien, Chung-Fang, Huang, Yu-Chen, Lin, Che-Hung
• Primary Examiner(s): CHANG, JON CARLTON
• Application Number:
  US12/193,827
• Publication Number:
  US 20090169107A1
• Time in Patent Office: 1,365 Days
• Field of Search: 382/284, 382/260, 382/264
• US Class Current: 382/260
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G06T 7/33 using feature-based methods
• CPC Class Codes: G06V 10/16 using multiple overlapping images; Image stitching

Abstract

An apparatus and a method of recognizing image feature pixels is related to the recognizing method including receiving an image having a plurality of pixels; subsequent filtering the pixels at least once by detecting the edge pixels of objects in the received image based on pixel values of the pixels, so as to obtain a plurality of candidate pixels; and performing multiple difference of Gaussian (DOG) blurring operations on the candidate pixels to find the candidate pixels having a relative extreme value as a plurality of feature pixels. Therefore, the processing time of recognizing image feature pixels may be reduced.

Claims

1. A method of recognizing image feature pixels, comprising:

receiving an image having a plurality of pixels;

subsequent filtering of the pixels at least once by detecting edge pixels of an object in the image based on pixel values of the pixels, so as to obtain a plurality of candidate pixels;

performing difference of Gaussian (DOG) blurring operations on the candidate pixels to find the candidate pixels each of which has a relative extreme value as a plurality of feature pixels;

stitching different images having the same feature pixels in the corresponding boundaries to form a stitched image; and

fading the pixels in the corresponding boundaries of the stitched image.

2. The method of recognizing image feature pixels as claimed in claim 1, wherein the step of subsequent filtering the pixels at least once is filtering the pixels through comparing a brightness value of each of the pixels with the brightness values of the adjacent pixels.

3. The method of recognizing image feature pixels as claimed in claim 1, wherein the step of subsequent filtering of the pixels at least once is filtering the pixels through comparing a chroma value of each of the pixels with the chroma values of the adjacent pixels.

4. The method of recognizing image feature pixels as claimed in claim 1, wherein the step of subsequent filtering the pixels at least once comprises:

performing a first filtering of the pixels with the pixels of the image;

performing a second filtering of the pixels with the pixels obtained from the first filtering; and

using the pixels obtained from the last filtering as the candidate pixels.

5. The method of recognizing image feature pixels as claimed in claim 1, wherein the step of subsequent filtering the pixels at least once comprises:

zooming out the image;

sequentially calculating a plurality of gradient values related to each of the pixels of the zoomed-out image;

obtaining the maximum gradient values among the gradient values related to each of the pixels;

comparing each of the maximum gradient values with a threshold value, so as to obtain the pixels corresponding to the maximum gradient values larger than the threshold value;

if the current filtering is the last one, using the obtained pixels as the candidate pixels; and

if the current filtering is not the last one, providing the obtained pixels for the next filtering.

6. The method of recognizing image feature pixels as claimed in claim 5, wherein the step of subsequent filtering the pixels at least once further comprises:

after comparing each of the maximum gradient values with the threshold value, performing exclusion procedure on adjacent pixels of the found pixels to keep one of the adjacent pixels, so as to obtain the nonadjacent pixels.

7. The method of recognizing image feature pixels as claimed in claim 1, wherein the step of subsequent filtering the pixels at least once comprises:

creating a window block by using each of the pixels as a central point;

calculating a mean of the pixel values in the window block;

comparing the mean with the pixel value of the corresponding central point to obtain the pixel as a candidate pixel;

if the current filtering is the last one, using the obtained pixels as the candidate pixels; and

if the current filtering is not the last one, providing the obtained pixels to the next filtering step.

8. The method of recognizing image feature pixels as claimed in claim 7, wherein the step of subsequent filtering the pixels at least once further comprises:

after making comparison according to the mean, performing exclusion procedure on the adjacent pixels to keep one of the adjacent pixels, so as to obtain the nonadjacent pixels.

9. The method of recognizing image feature pixels as claimed in claim 1, further comprising:

selecting a color channel of the image to filter the following pixels corresponding to the color channel.

10. An apparatus of recognizing image feature pixels, comprising:

an image receiving end, for receiving an image having a plurality of pixels;

at least one filter, which is sequentially connected in series, wherein a filter in the most upstream among the filters is electrically connected to the image receiving end, and the filter is used for subsequent filtering of the pixels to detect edge of objects in the receiving image based on pixel values of the pixels, so as to obtain a plurality of candidate pixels, wherein at least one of the filter filters is a gradient filter, and each gradient filter comprises;

an image input end, for receiving the image from the upstream filter;

a zoom unit, electrically connected to the image input end, for zooming out the image received by the image input end;

a detection unit, electrically connected to the zoom unit, for sequentially calculating a plurality of gradient values related to each of the pixels on the zoomed-out image and outputting the maximum gradient values among the gradient values related to each of the pixels;

a determination unit, electrically connected to the detection unit, for comparing each of the maximum gradient values with a threshold value and marking the pixels corresponding to the maximum gradient values larger than the threshold value; and

an image output end, electrically connected to the determination unit, for outputting the image having the pixels marked by the determination unit;

wherein when the image from the upstream has the marked pixels, the detection unit calculates the gradient values for the marked pixels, and the determination unit remarks the pixels after comparison; and

a feature describing module, electrically connected to the image receiving end and the filter in the most downstream among the filters, for performing difference of Gaussian (DOG) blurring operations based on the candidate pixels output from the filter in the most downstream to find the pixels each of which has a relative extreme value as a plurality of feature pixels.

11. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein in at least one filter among the filters, the filter filtering the pixels based on brightness values of the pixels.

12. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein in at least one filter among the filters, the filter filtering the pixels based on chroma values of the pixels.

13. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein and each of the gradient filters further comprises:

an exclusion unit, electrically connected to the determination unit, for performing exclusion procedure on adjacent pixels marked by the determination unit, so as to keep one of the adjacent pixels;

wherein the image output end electrically connects to the exclusion unit and electrically connects to the exclusion unit via the exclusion unit, for outputting the image having the marked pixels kept by the exclusion unit.

14. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein at least one filter among the filters is a mean filter, and each of the mean filters comprises:

an image input end, for receiving the image from the upstream filter;

a window creation unit, electrically connected to the image input end, for creating a window block by using each of the pixels in the image received by the image input end as a central point;

a mean calculation unit, electrically connected to the window creation unit, for calculating a mean of the pixel values in the window block;

a determination unit, electrically connected to the mean calculation unit and the window creation unit, for marking the pixels with the pixel values having a difference with respect to the mean; and

an image output end, electrically connected to the determination unit, for outputting the image having the pixels marked by the determination unit;

wherein the image from the upstream has the marked pixels, the window creation unit creates the window block for the marked pixels, and the determination unit remarks the pixels of the image according to the marked pixels.

15. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein at least filter among the filters is a mean filter, and each of the mean filters comprises:

an image input end, for receiving the image from the upstream filter;

a window creation unit, electrically connected to the image input end, for creating a window block by using each of the pixels in the image received by the image input end as a central point;

a mean calculation unit, electrically connected to the window creation unit, for calculating a mean of the pixel values in the window block;

a determination unit, electrically connected to the mean calculation unit and the window creation unit, for marking the pixels with the pixel values having a difference with respect to the mean;

an exclusion unit, electrically connected to the determination unit, for performing exclusion procedure on adjacent pixels in the pixels marked by the determination unit, so as to keep one of the adjacent pixels; and

an image output end, electrically connected to the exclusion unit, for outputting the image having the marked pixels kept by the exclusion unit;

wherein the image from the upstream has the marked pixels, the window creation unit creates the window block for the marked pixels, and the determination unit remark the pixels of the image according to the found pixels.

16. The apparatus of recognizing image feature pixels as claimed in claim 10, wherein the image receiving end comprises:

a channel selecting module, electrically connected to the filter in the most upstream among the filters and the feature describing module, for selecting a color channel of the received image and outputting the image corresponding to the color channel.