improve performance of {invert, solarize}_image_tensor (#6819)

* improve performance of invert_image_tensor

* cleanup

* lint

* more cleanup

* use new invert in solarize

torchvision/prototype/transforms/functional/_color.py

@@ -312,7 +312,13 @@ def posterize(inpt: features.InputTypeJIT, bits: int) -> features.InputTypeJIT:
         return posterize_image_pil(inpt, bits=bits)
 
 
-solarize_image_tensor = _FT.solarize
+def solarize_image_tensor(image: torch.Tensor, threshold: float) -> torch.Tensor:
+    if threshold > _FT._max_value(image.dtype):
+        raise TypeError(f"Threshold should be less or equal the maximum value of the dtype, but got {threshold}")
+
+    return torch.where(image >= threshold, invert_image_tensor(image), image)
+
+
 solarize_image_pil = _FP.solarize
 
 
@@ -456,7 +462,13 @@ def equalize(inpt: features.InputTypeJIT) -> features.InputTypeJIT:
         return equalize_image_pil(inpt)
 
 
-invert_image_tensor = _FT.invert
+def invert_image_tensor(image: torch.Tensor) -> torch.Tensor:
+    if image.dtype == torch.uint8:
+        return image.bitwise_not()
+    else:
+        return _FT._max_value(image.dtype) - image  # type: ignore[no-any-return]
+
+
 invert_image_pil = _FP.invert