diff --git a/index.html b/index.html index 7c0a759..bff6fe0 100644 --- a/index.html +++ b/index.html @@ -166,7 +166,7 @@

Deblurring 3D Gaussian Splatting

Abstract

- +

@@ -192,10 +192,13 @@

Abstract

Deblurring 3D Gaussians

- +
+ +

We learn the deblurring by transforming the geometry of the 3D Gaussians. To do so, we have employed an MLP that takes the position, rotation, scale, and viewing direction of 3D Gaussians as inputs, and outputs offsets for rotation and scale. Then these offsets are element-wisely multiplied to rotation and scale, respectively, to obtain the transformed geometry of the 3D Gaussians.

+

Since we predict the offsets for each Gaussian, we can selectively enlarge the covariances of Gaussians where the parts in the training images are blurred. This flexibility enables us to effectively implement deblurring capability in 3D-GS. On the other hand, a naive approach to blurring the rendered image is simply to apply a Gaussian kernel which is not capable of handling each part of the image differently but blurs @@ -249,6 +252,59 @@

Compensation for Sparse Point Cloud

+
+
+ +
+
+

Visualization of Defocus and Camera Motion blur

+
+

Gaussians visualization for defocus blur

+
+ + +

This figure visualizes the original and transformed 3D Gaussians for + defocus blur. With a given view whose near plane is defocused, the transformed + 3D Gaussians show larger scales than those of the original 3D Gaussians to + model defocus blur on the near plane (blur-bordered images). Meanwhile, the + transformed 3D Gaussians keep very similar shapes to the original ones for sharp + objects in the far plane (red-bordered images). +

+
+
+

Point cloud visualization for camera motion blur

+
+ + +

This figure depicts point clouds of the original 3D Gaussians and transformed 3D Gaussians. + The point cloud of the transformed 3D Gaussians exhibits camera movements when the camera moves left to right. +

+
+
+
+
+
+ +
+
+ +
+
+

Selective Gaussian blur adjustment

+ + +

Selective Gaussian blur adjustment. As delineated in this figure, our methodology + adeptly harnesses the δrj , δsj both emanating from compact Multi-Layer Perceptrons + (MLP), enabling the inversion of Gaussian blur regions or the comprehensive modulation of overall blurriness and sharpness. + With the Transformation of δrj , δsj , our framework facilitates the nuanced blurring of proximal regions akin to A, as well as + the deft blurring of distant locales akin to B. Furthermore, it offers the capability to + manipulate the global blurriness or sharpness, exemplified by adjustments akin to C and D. +

+
+
+
+
+
diff --git a/static/images/FPS_curve.jpg b/static/images/FPS_curve.jpg new file mode 100644 index 0000000..5a100e0 Binary files /dev/null and b/static/images/FPS_curve.jpg differ diff --git a/static/images/FPS_curve.png b/static/images/FPS_curve.png deleted file mode 100644 index 4f8c832..0000000 Binary files a/static/images/FPS_curve.png and /dev/null differ diff --git a/static/images/gaussian_visualization.jpg b/static/images/gaussian_visualization.jpg new file mode 100644 index 0000000..32107a2 Binary files /dev/null and b/static/images/gaussian_visualization.jpg differ diff --git a/static/images/pointcloud_visualization.jpg b/static/images/pointcloud_visualization.jpg new file mode 100644 index 0000000..2c81bcc Binary files /dev/null and b/static/images/pointcloud_visualization.jpg differ diff --git a/static/images/selective_figure_ext.jpg b/static/images/selective_figure_ext.jpg new file mode 100644 index 0000000..b25de3b Binary files /dev/null and b/static/images/selective_figure_ext.jpg differ diff --git a/static/images/workflow.png b/static/images/workflow.png index b31d87f..53b0de8 100644 Binary files a/static/images/workflow.png and b/static/images/workflow.png differ