Method for detecting shifts in line images obtained by a sensor that is airborne or moving in space
Method for estimating shifts of line images obtained by an airborne or spaceborne scanning sensor. The invention relates to a method for estimating shifts in line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images. The line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over the celestial body. The shifts in the line images are induced by sighting errors of the sensor ( 10 ) during the acquisition of the line image L j . The lateral and/or longitudinal shift of a line image L j relative to a line image L i , where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images.
1. An estimation method for estimating shifts of line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images, in which:
the line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over said celestial body, each line image being composed of a plurality of pixels, and
a line image L j , where 1≦j≦J, presents a lateral and/or longitudinal shift between the position that line image L j is destined to occupy in the composite image and the position it should occupy, because of a sighting error of the sensor ( 10 ) during the acquisition of line image L j ,
wherein the lateral and/or longitudinal shift of a line image L j relative to a line image L i , where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images,
wherein the local model of lateral and/or longitudinal variation of the similarity between different line images is determined by calculating similarity values between different line images of the composite image,
a step ( 20 ) of estimating the relative lateral shift δ j of line image L j relative to line image L i , during which the local model of lateral variation of the similarity between successive line images is determined by calculating similarity values of line image L j to line image L i for a plurality of lateral translation values of one line relative to the other,
the relative lateral shift δ j is estimated as the lateral translation for which the value of the model of lateral variation of the similarity is maximum, and
a step ( 21 ) of integrating relative lateral shifts δ j in successive line images, for determining cumulative lateral shifts δ′ j with respect to a reference line image of the composite image.
2. The estimation method according to claim 1 , comprising a step ( 22 ) of low-cut filtering of relative lateral shifts δ j or cumulative lateral shifts δ′ j .
3. An estimation method for estimating shifts of line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images, in which:
the line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over said celestial body, each line image being composed of a plurality of pixels, and
a line image L j , where 1≦j≦J, presents a lateral and/or longitudinal shift between the position that line image L j is destined to occupy in the composite image and the position it should occupy, because of a sighting error of the sensor ( 10 ) during the acquisition of line image L j ,
wherein the lateral and/or longitudinal shift of a line image L j relative to a line image L i , where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images,
a step ( 23 ) of estimating the relative longitudinal shift Δ j of line image L j relative to line image L i , during which a similarity value is calculated for line image L j and line image L i , and the relative longitudinal shift Δ j is estimated according to a longitudinal gap corresponding, based on the local model of longitudinal variation of the similarity, to said similarity value for line image L j and line image L i ,
wherein when lateral shifts have already been estimated, the step ( 23 ) of estimating the relative longitudinal shift Δ j is applied to line images L′ j instead of line images L j , where 1≦j≦J, said line images L′ j corresponding to lines images L j after said estimated lateral shifts have been corrected.
4. An estimation method for estimating shifts of line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images, in which:
the line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over said celestial body, each line image being composed of a plurality of pixels, and
a line image L j , where 1≦j≦J, presents a lateral and/or longitudinal shift between the position that line image L j is destined to occupy in the composite image and the position it should occupy, because of a sighting error of the sensor ( 10 ) during the acquisition of line image L j ,
wherein the lateral and/or longitudinal shift of a line image L j relative to a line image L i , where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images,
a step ( 23 ) of estimating the relative longitudinal shift Δ j of line image L j relative to line image L i , during which a similarity value is calculated for line image L j and line image L i , and the relative longitudinal shift Δ j is estimated according to a longitudinal gap corresponding, based on the local model of longitudinal variation of the similarity, to said similarity value for line image L j and line image L i
wherein the local model of longitudinal variation of the similarity between successive line images is calculated, for a given theoretical longitudinal gap between two line images, by averaging similarity values calculated for a plurality of pairs of line images of the composite image separated by said given theoretical longitudinal gap.
5. The estimation method according to claim 4 , in which the local model of longitudinal variation of the similarity between successive line images is calculated between the theoretical longitudinal gaps by interpolation according to values calculated for the theoretical longitudinal gaps.
6. An estimation method for estimating shifts of line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images, in which:
the line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over said celestial body, each line image being composed of a plurality of pixels, and
a line image L j , where 1≦j≦J, presents a lateral and/or longitudinal shift between the position that line image L j is destined to occupy in the composite image and the position it should occupy, because of a sighting error of the sensor ( 10 ) during the acquisition of line image L j ,
wherein the lateral and/or longitudinal shift of a line image L j relative to a line image L i , where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images,
a step ( 23 ) of estimating the relative longitudinal shift Δ j of line image L j relative to line image L i , during which a similarity value is calculated for line image L j and line image L i , and the relative longitudinal shift Δ j is estimated according to a longitudinal gap corresponding, based on the local model of longitudinal variation of the similarity, to said similarity value for line image L j and line image L i , and
a step ( 24 ) of integrating relative longitudinal shifts Δ j in successive line images for determining cumulative longitudinal shifts Δ′ j with respect to a reference line image of the composite image.
7. The estimation method according to claim 6 , comprising a step ( 25 ) of low-cut filtering of relative or cumulative longitudinal shifts.
8. A method for forming composite images using line images with shifts, comprising:
a step ( 30 ) of estimating shifts of a first set of line images acquired by a first sensor ( 10 ), utilizing an estimation method for estimating shifts of line images of an image, referred to as a “composite image”, made up of the juxtaposition of J line images, in which:
the line images show portions of a scene on the surface (S) of a celestial body that are different and successively acquired by a row of sensitive cells of a sensor ( 10 ) on board a vehicle flying over said celestial body, each line image being composed of a plurality of pixels, and
a line image Lj, where 1≦j≦J, presents a lateral and/or longitudinal shift between the position that line image Lj is destined to occupy in the composite image and the position it should occupy, because of a sighting error of the sensor ( 10 ) during the acquisition of line image Lj,
wherein the lateral and/or longitudinal shift of a line image Lj relative to a line image Li, where j is different from i, is estimated according to a local model of lateral and/or longitudinal variation of the similarity between different line images,
a step of converting estimated shifts in a second set of shifts for a second set of line images, acquired by a second sensor on board the vehicle and substantially subjected to the same sighting errors as the first sensor ( 10 ),
a step ( 31 ) of correcting the second set of shifts of the second set of line images, and
a step ( 32 ) of juxtaposing the second set of corrected line images.