Sobel edge extraction circuit for image processing
First Claim
Patent Images
1. An image processing device on a single integrated circuit chip for performing the Sobel algorithm S={[(a+2b+c)-(g+2f+e)]2 +[(a+2h+g)-(c+2d+e)]2 }1/2 with digitized input data words derived from an image by use of a 3×
- 3 window in the form
space="preserve" listing-type="tabular">______________________________________ a b c h z d g f e ______________________________________ supplied via input data bus means, said device comprising;
input logic means including input register means and parallel adder means for performing the functions J=(a+2b+c), K=(g+2f+e), L=(a+2h+g), M=(c+2d+e), P=|J-K| and Q=|L-M| which produces words P and Q, the input register means being coupled to said input data bus means;
first memory means storing binary words representing the squares of the binary numbers 0 to 111 1111 with 13 bits per word for the squares using 7-bit addresses, first selection means for selecting the seven least significant bits of each of said words P and Q if the remaining more significant bits are all zero'"'"'s and otherwise selecting the binary number 111 1111 as addresses to produce memory outputs from said first memory means for P2 and Q2, register means coupled to the first memory means output and adder means coupled thereto to provide a 13-bit word for the value V=(P2 +Q2);
second memory means storing binary words of six bits each representing the square root of V, comprising a first memory section of 1024 words with addresses formed from the ten least significant bits of V, and a second memory section with addresses formed from the three most significant bits of V, with second selection means for selecting an output from said first section if the three most significant bits are all zero'"'"'s and otherwise to select an output from said second section, and means to supply the selected output as the Sobel square root value S.
1 Assignment
0 Petitions
Accused Products
Abstract
The Sobel square root algorithm S={[(a+2b+c)-(g+2f+e)]2 +[(a+2H+g)-(c+2d+e)]2 }1/2, with 8-bit input data from a 3×3 window and 6-bit output is performed on a single VLSI chip, using a square table only 128×13 and a square root table only 1027 or 1032×6 in ROM. The random logic including adders and clock circuits are also on the same chip with the ROM tables.
23 Citations
7 Claims
-
1. An image processing device on a single integrated circuit chip for performing the Sobel algorithm S={[(a+2b+c)-(g+2f+e)]2 +[(a+2h+g)-(c+2d+e)]2 }1/2 with digitized input data words derived from an image by use of a 3×
- 3 window in the form
space="preserve" listing-type="tabular">______________________________________ a b c h z d g f e ______________________________________supplied via input data bus means, said device comprising; input logic means including input register means and parallel adder means for performing the functions J=(a+2b+c), K=(g+2f+e), L=(a+2h+g), M=(c+2d+e), P=|J-K| and Q=|L-M| which produces words P and Q, the input register means being coupled to said input data bus means; first memory means storing binary words representing the squares of the binary numbers 0 to 111 1111 with 13 bits per word for the squares using 7-bit addresses, first selection means for selecting the seven least significant bits of each of said words P and Q if the remaining more significant bits are all zero'"'"'s and otherwise selecting the binary number 111 1111 as addresses to produce memory outputs from said first memory means for P2 and Q2, register means coupled to the first memory means output and adder means coupled thereto to provide a 13-bit word for the value V=(P2 +Q2); second memory means storing binary words of six bits each representing the square root of V, comprising a first memory section of 1024 words with addresses formed from the ten least significant bits of V, and a second memory section with addresses formed from the three most significant bits of V, with second selection means for selecting an output from said first section if the three most significant bits are all zero'"'"'s and otherwise to select an output from said second section, and means to supply the selected output as the Sobel square root value S. - View Dependent Claims (2, 3, 4)
- 3 window in the form
-
5. A signal processing device comprising:
-
input logic means which during each of a sequence of time intervals produces two parallel digital data words P and Q, each having a given number of bits greater than N; first memory means storing binary words representing the squares of the 2N binary numbers 0 to (2N -1) with at least (2N-1) bits per word for the squares using addresses of N bits, first selection means for selecting the N least significant bits of each of said words P and Q if the remaining more significant bits are all zero'"'"'s and otherwise selecting the binary number (2N -1) as addresses to produce memory outputs from said first memory means for P2 and Q2, register means coupled to the first memory means output and adder means coupled thereto to provide a word having at least (2N-1) bits for the value V=(P2 +Q2); second memory means storing binary words representing the square root of V of approximately half the number of bits as words in the first memory means, the second memory means being comprised of first and second sections, the first section having addresses formed from a predetermined number of the least significant bits of words V, and the second section having addresses formed from the remaining more significant bits of V, with second selection means for selecting an output from said first section if said remaining more significant bits of V are all zero'"'"'s and otherwise to select an output from said second section, and means to supply the selected output as the output of said device representing √
P2 +Q2 during each of said time intervals. - View Dependent Claims (6, 7)
-
Specification