Image processing method for the detection of the horizon and device for the implementation thereof

• US 5,214,720 A
• Filed: 07/19/1991
• Issued: 05/25/1993
• Est. Priority Date: 07/27/1990
• Status: Expired due to Fees

• Current Assignee: Thomson Trt Defense
• Original Assignee: Thomson Trt Defense
• Inventors: Fety, Luc, Terre, Michel, Noreve, Xavier
• Primary Examiner(s): Couso, Jose L.
• Application Number:
  US07/732,917
• Time in Patent Office: 676 Days
• Field of Search: 358/103, 358/105, 358/108, 358/109, 358/113, 358/160, 382/1, 382/30, 382/34, 382/48
• US Class Current: 382/100
• CPC Class Codes: G01S 3/7868 using horizon sensors
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/12 Edge-based segmentation

Abstract

According to the method, after a step for the digital processing of the video signal of the image analyzed by rows, there is determined, for each row, on the basis of the distribution of the mean light energy values below and above each position of the row, a step modelling function, the jump position of which gives the position of the horizon in this row. The device used to implement this method includes a sampler of the video signal coupled with a flip flop memory through an accumulator which adds up the luminance values row by row, and a processor which, on the basis of the distribution of the mean light energy levels, computes the signal to be subtracted from the digital video signal by means of a subtractor. The disclosed method and device can be applied to the detection of the horizon from an image analyzed by columns or by lines.

Claims

1. An image processing method fcr the detection of the horizon from the video signal de1ivered by a camera forming the image of the observation field analyzed row by row, wherein said image processing method comprises the following steps:

the digital processing of the video signal to obtain the sampled luminance values of the analyzed image row by row;

the determination, for each position i of a given row k, of a step modelling function, the jump of which is located at i and the amplitude of which is computed so as to give a minimal error of all the values assumed by the function with respect to all the values of the luminance signal of the row considered;

the selection of the value i_hk of i, corresponding to the position of the horizon on the row considered, for which the step function determined at the previous step has an amplitude computed so as to give the smallest of all the minimal errors;

the reproduction of the preceding steps for all the rows of the analyzed image and the display of a smoothing curve, representing the horizon, from the set of the value i_hk sselected.

2. An image processing method for the detection of the horizon according to claim 1, wherein the step modelling function is the sum of twc decorrelated step functions, the amplitudes of which have values such that, for each position i of a given row, the error between the values of the modelling function and the set of the sampled values, taken respectively at the corresponding level by the luminance signal, is minimized.

3. An image processing method according to either of the claims 1 or 2, wherein the level i_hk of the position of the horizon on the order k row is that for which the sum of the mean light levels below and above this level is maximal in the row considered.

4. An image processing method according to either of the claims 1 or 2, applied to a columnwise analysis of images, wherein the image is cut up into p columns, wherein i varies linewise from 1 to n and wherein the selected values i_hk directly indicate the height of the horizon on the image.

5. An image processing method according to either of the claims 1 or 2, applied to a linewise analysis of images, wherein the image is cut up into n lines, wherein i varies columnwise from 1 to p and wherein the selected values i_hk either directly or indirectly indicate, after the processing of the luminance values by columns, the height of the horizcn on the image.

6. An image processing device for the detection of the horizon from the video signal delivered by a camera forming the image of the observation field analyzed row by row, wherein the video signal is digitablly processed to obtain sampled luminance values of the analyzed image row by row;

for each position i of a given row k, a step modelling function is determined, the jump of which is located at ia dn the amplitdue of which is computed so as to give a minimal error of all the values assumed by the function with respect to all the values of the luminance signal of the row considered, a value of i_hk of i, corresponding to the position of the horizon on the row considered, is selected for which the determined step function has an amplitude computer so as to give the smallest of all the minimal errors, the value i_hk for all the rows of the analyzed image is determined and s moothing curve, representing the horizon, from the set of values i_hk selected is displayed, wherein the determination of said step modelling function is performed using a columnwise analysis in which the image is cut up into p columns, i varies linewise from 1 to n, and the selected values i_hk directly indicate the height of the horizon on the image, said device comprising;

a signal sampler (b), the digital values of which are successively addressed to an accumulator to add up these digitized values as when they appear, a flip flop memory that is coupled to the accumulator and memorizes the cumulated values, a signal processor which, on the basis of the values transmitted by the memory and for each position i of each column k of the luminance values, determines the quantity q_i equal to the sum of the mean light energies above and below the position i, a comparator coupled to the processor to select the quantity q_Mk having the maximum value for the order k column, memories coupled to the comparator to respectively memorize the quantities q_Mk and the corresponding value i_hk of i, and a subtractor mounted on the circuit delivering the sample video signal and on the processor to subtract, from the signal coming from the sampler, a maximum luminance signal coming from the processor for the level of each column in which the point of the straight line, representing the position of the horizon on the column considered, is positioned, this straight line being computed in the processor by linear regression by application of the values of levels memorized in the memory and of the corresponding weighting values memorized in the memory.

7. An image processing device for the detection of the horizon from the video signal delivered by a camera forming the image of the observation field analyzed row by row, wherein the video signal is digitally processed to obtain sampled luminance values of the analyzed image row by row;

for each position i of a given row k, a step modelling function is determined, the jump of which is located at i and the amplitude of which is computed so as to give a minimal error of all the values assumed by the function with respect to all the values of the luminance signal of the row considered, a value of i_hk of i, corresponding to the position of the horizon on the row considered, is selected for which the determined ste function has an amplitude computed so as to give the smallest of all the minimal errors, the value i_hk for all the rows of the analyzed image is determined and a smoothing curve, representing the horizon, from the set of values i_hk selected is displayed, wherein the determination of said step modelling function is performed using a linewise analysis in which the image is cut up into n lines, i varies linewise from 1 to p, and the selected values i_hk either directly or indirectly indicate, after processing the luminance values by column, the height of the horizon on the image, said device comprising;

a signal sampler (b), the digital values of which are successively addressed to an accumulator to add up these digitized values as and when they appear, a pair of flip flop memories that is coupled to the accumulator and memorizes the cumulated values, a signal processor which, on the basis of the values transmitted by the memory and for each position i of each column k of the luminance values, determines the quantity q_i equal to the sum of the mean light energies above and below the position i, a comparator coupled to the processor to select a quantity q_Mk having the maximum value for the order k column, memories coupled to the comparator to respectively memorize the quantities q_Mk and the corresponding values i_hk of i, and a subtractor mounted on the circuit delivering the sampled video signal and on the processor to subtract, from the signal coming from the sampler, a maximum luminance signal coming from the processor for the level of each column in which the point of the straight line, representing the position of the horizon on the column considered, is positioned, this straight line being computed in the processor by linear regression by application of the values of levels memorized in the emmory and of the corresponding weighting values memorized in the memory, wherein a first of the flip flop memories successively memorizes the partial summing operations of the luminance values to cumulate and store the luminance values column by column and wherein a second of the flip flop memories stores the luminance values of the columns of the previous image frame, wherein the quantities q_i are computed in the processor from the partial summing operations of the luminance values of the columns coming from the first of the flip flop memories and from the summing operations of the luminance values of the columns coming from the second of the flip flop memories.