| Safe Haskell | None |
|---|---|
| Language | Haskell2010 |
OpenCV.Photo
- data InpaintingMethod
- decolor :: Mat (S [h, w]) (S 3) (S Word8) -> CvExcept (Mat (S [h, w]) (S 1) (S Word8), Mat (S [h, w]) (S 3) (S Word8))
- inpaint :: channels `In` [1, 3] => Double -> InpaintingMethod -> Mat (S [h, w]) (S channels) (S Word8) -> Mat (S [h, w]) (S 1) (S Word8) -> CvExcept (Mat (S [h, w]) (S channels) (S Word8))
- denoise_TVL1 :: Double -> Int32 -> Vector (Mat (S [h, w]) (S 1) (S Word8)) -> CvExcept (Mat (S [h, w]) (S 1) (S Word8))
- fastNlMeansDenoisingColored :: Double -> Double -> Int32 -> Int32 -> Mat (S [h, w]) (S 3) (S Word8) -> CvExcept (Mat (S [h, w]) (S 3) (S Word8))
- fastNlMeansDenoisingColoredMulti :: Double -> Double -> Int32 -> Int32 -> Vector (Mat (S [h, w]) (S 3) (S Word8)) -> CvExcept (Mat (S [h, w]) (S 3) (S Word8))
Documentation
data InpaintingMethod Source #
Constructors
| InpaintNavierStokes | Navier-Stokes based method. |
| InpaintTelea | Method by Alexandru Telea. |
Instances
Arguments
| :: Mat (S [h, w]) (S 3) (S Word8) | Input image. |
| -> CvExcept (Mat (S [h, w]) (S 1) (S Word8), Mat (S [h, w]) (S 3) (S Word8)) | Output images. |
Perform decolor
Decolor a color image to a grayscale (1 channel) and a color boosted image (3 channel)
Example:
decolorImg
:: forall h h2 w w2 c d
. ( Mat (ShapeT [h, w]) ('S c) ('S d) ~ Kodak_512x341
, h2 ~ ((*) h 2)
, w2 ~ ((*) w 2)
)
=> Mat ('S ['S h2, 'S w2]) ('S c) ('S d)
decolorImg = exceptError $ do
(bikesGray, boost) <- decolor bikes_512x341
colorGray <- cvtColor gray bgr bikesGray
withMatM
(Proxy :: Proxy [h2, w2])
(Proxy :: Proxy c)
(Proxy :: Proxy d)
white $ \imgM -> do
matCopyToM imgM (V2 0 0) bikes_512x341 Nothing
matCopyToM imgM (V2 0 h) colorGray Nothing
matCopyToM imgM (V2 w h) boost Nothing
where
w = fromInteger $ natVal (Proxy :: Proxy w)
h = fromInteger $ natVal (Proxy :: Proxy h)
Arguments
| :: channels `In` [1, 3] | |
| => Double | inpaintRadius - Radius of a circular neighborhood of each point inpainted that is considered by the algorithm. |
| -> InpaintingMethod | |
| -> Mat (S [h, w]) (S channels) (S Word8) | Input image. |
| -> Mat (S [h, w]) (S 1) (S Word8) | Inpainting mask. |
| -> CvExcept (Mat (S [h, w]) (S channels) (S Word8)) | Output image. |
Restores the selected region in an image using the region neighborhood.
Example:
inpaintImg
:: forall h h2 w w2 c d
. ( Mat (ShapeT [h, w]) ('S c) ('S d) ~ Kodak_512x341
, h2 ~ ((*) h 2)
, w2 ~ ((*) w 2)
)
=> Mat ('S ['S h2, 'S w2]) ('S c) ('S d)
inpaintImg = exceptError $ do
maskInv <- bitwiseNot mask
maskBgr <- cvtColor gray bgr maskInv
damaged <- bitwiseAnd bikes_512x341 maskBgr
repairedNS <- inpaint 3 InpaintNavierStokes damaged mask
repairedT <- inpaint 3 InpaintTelea damaged mask
withMatM
(Proxy :: Proxy [h2, w2])
(Proxy :: Proxy c)
(Proxy :: Proxy d)
black $ \imgM -> do
matCopyToM imgM (V2 0 0) damaged Nothing
matCopyToM imgM (V2 w 0) maskBgr Nothing
matCopyToM imgM (V2 0 h) repairedNS Nothing
matCopyToM imgM (V2 w h) repairedT Nothing
where
mask = damageMask
w = fromInteger $ natVal (Proxy :: Proxy w)
h = fromInteger $ natVal (Proxy :: Proxy h)
Arguments
| :: Double | details more is more 2 |
| -> Int32 | Number of iterations that the algorithm will run |
| -> Vector (Mat (S [h, w]) (S 1) (S Word8)) | Vector of odd number of input 8-bit 3-channel images. |
| -> CvExcept (Mat (S [h, w]) (S 1) (S Word8)) | Output image same size and type as input. |
Perform denoise_TVL1
Example:
denoise_TVL1Img
:: forall h w w2 c d
. ( Mat (ShapeT [h, w]) ('S c) ('S d) ~ Lenna_512x512
, w2 ~ ((*) w 2)
)
=> Mat ('S ['S h, 'S w2]) ('S c) ('S d)
denoise_TVL1Img = exceptError $ do
denoised <- matChannelMapM (denoise_TVL1 2 50 . V.singleton) lenna_512x512
withMatM
(Proxy :: Proxy [h, w2])
(Proxy :: Proxy c)
(Proxy :: Proxy d)
black $ \imgM -> do
matCopyToM imgM (V2 0 0) lenna_512x512 Nothing
matCopyToM imgM (V2 w 0) denoised Nothing
where
w = fromInteger $ natVal (Proxy :: Proxy w)
fastNlMeansDenoisingColored Source #
Arguments
| :: Double | Parameter regulating filter strength for luminance component. Bigger h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise |
| -> Double | The same as h but for color components. For most images value equals 10 will be enough to remove colored noise and do not distort colors |
| -> Int32 | templateWindowSize Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels |
| -> Int32 | searchWindowSize. Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater searchWindowsSize - greater denoising time. Recommended value 21 pixels |
| -> Mat (S [h, w]) (S 3) (S Word8) | Input image 8-bit 3-channel image. |
| -> CvExcept (Mat (S [h, w]) (S 3) (S Word8)) | Output image same size and type as input. |
Perform fastNlMeansDenoising function for colored images. Denoising is not per channel but in a different colour space
Example:
fastNlMeansDenoisingColoredImg
:: forall h w w2 c d
. ( Mat (ShapeT [h, w]) ('S c) ('S d) ~ Lenna_512x512
, w2 ~ ((*) w 2)
)
=> Mat ('S ['S h, 'S w2]) ('S c) ('S d)
fastNlMeansDenoisingColoredImg = exceptError $ do
denoised <- fastNlMeansDenoisingColored 3 10 7 21 lenna_512x512
withMatM
(Proxy :: Proxy [h, w2])
(Proxy :: Proxy c)
(Proxy :: Proxy d)
black $ \imgM -> do
matCopyToM imgM (V2 0 0) lenna_512x512 Nothing
matCopyToM imgM (V2 w 0) denoised Nothing
where
w = fromInteger $ natVal (Proxy :: Proxy w)
fastNlMeansDenoisingColoredMulti Source #
Arguments
| :: Double | Parameter regulating filter strength for luminance component. Bigger h value perfectly removes noise but also removes image details, smaller h value preserves details but also preserves some noise |
| -> Double | The same as h but for color components. For most images value equals 10 will be enough to remove colored noise and do not distort colors |
| -> Int32 | templateWindowSize Size in pixels of the template patch that is used to compute weights. Should be odd. Recommended value 7 pixels |
| -> Int32 | searchWindowSize. Size in pixels of the window that is used to compute weighted average for given pixel. Should be odd. Affect performance linearly: greater searchWindowsSize - greater denoising time. Recommended value 21 pixels |
| -> Vector (Mat (S [h, w]) (S 3) (S Word8)) | Vector of odd number of input 8-bit 3-channel images. |
| -> CvExcept (Mat (S [h, w]) (S 3) (S Word8)) | Output image same size and type as input. |
Perform fastNlMeansDenoisingColoredMulti function for colored images. Denoising is not pre channel but in a different colour space. This wrapper differs from the original OpenCV version by using all input images and denoising the middle one. The original version would allow to have some arbitrary length vector and slide window over it. As we have to copy the haskell vector before we can use it as `std::vector` on the cpp side it is easier to trim the vector before sending and use all frames.
Example:
fastNlMeansDenoisingColoredMultiImg
:: forall h w w2 c d
. ( Mat (ShapeT [h, w]) ('S c) ('S d) ~ Lenna_512x512
, w2 ~ ((*) w 2)
)
=> Mat ('S ['S h, 'S w2]) ('S c) ('S d)
fastNlMeansDenoisingColoredMultiImg = exceptError $ do
denoised <- fastNlMeansDenoisingColoredMulti 3 10 7 21 (V.singleton lenna_512x512)
withMatM
(Proxy :: Proxy [h, w2])
(Proxy :: Proxy c)
(Proxy :: Proxy d)
black $ \imgM -> do
matCopyToM imgM (V2 0 0) lenna_512x512 Nothing
matCopyToM imgM (V2 w 0) denoised Nothing
where
w = fromInteger $ natVal (Proxy :: Proxy w)
