CS194-26 Final Project

Anusha Syed and Imran Jami

Fake Miniatures

Overview

The general idea of this project was to create a miniature setting with normal everyday photos. This was a really cool part of the project, and was really enjoyable to see our results come to life!

Our approach was basically to take the normal photo, and a mask of where we would want to concentrate our focus on for this photo, and blend the image with the mask in order to crete a custom focus in the image so that the "miniature" effect.

Implementation

Our general Implementation was as follows: 1. Load all images/masks, including a blurred version of the original image 2. Generate a laplacian stack for both the original and blurred image & the mask 3. Take both images, and apply the mask on each image 4. Combine both masked images together to create our final image 5. Edit color and contrast of the image to enhance the miniature affect

Results

Stock photo example:

Original Photo

Miniature Photo

Example of Santa Monica with Vertcal Mask:

Original Photo

Miniature Photo

Example of Disneyland with Horizontal Mask:

Original Photo

Miniature Photo

Example of San Francisco with Horizontal Mask:

Original Photo

Miniature Photo

Seam Carving

Overview

The general idea of this project was to Implement a Seam Carving algorithm in order to resize images without losing any major features of the image. An example of this would be a large landscape of the ocean with one boat. In theory we would be able to shrink the image by cropping out different vertical seams of the image that are very similar to each other.

We followed a Dynamic Programming Algorithm in order to achieve this effect with existing images. We generate different vertical and horizontal seams of the image, and take out the pixels of the image that have little importance to the overall visual that is involved in the photograph.

Implementation

The general approach was to do this for both the verical and horizontal seams of the image. We first generated the importance of each pixel with our energy function, and then removed the lowest importance seam of the image in each iteration until we were at the proper desired dimension for the resulting image. This worked fairly well, and here are some of our results.

Results

Horizontal cropping

Original Photo

Energy Function Result

Cropped Photo

Original Photo

Cropped Photo

Original Photo

Cropped Photo

Original Photo

Cropped Photo

Vertical cropping

Original Photo

Cropped Photo

Original Photo

Cropped Photo

Failed Cases:

Original

Cropped Photo

Original

Cropped Photo

Vertigo Shot

Overview

This part of the project was pretty cool, because we got to recreate famous photographic techniques used in movies that everyone has seen. The Vertigo Shot was also pretty fun to recreate.

The process involves taking a series of photos of the same subject at different FOV during each step. Our results were pretty good, and were similar to the famous Vertigo Shot.

Implementation

The implementation was pretty straight forward. We took a series of photos of the same subject at different field of views with a camera. As we took the photos we got closer and close to the subject to create this effect.

Results

Dolly Potion 1

Dolly Potion 2

Dolly Potion 3

Dolly Potion 4

Dolly Potion

Dolly Tiger

Website design template from https://inst.eecs.berkeley.edu/~cs194-26/fa16/upload/files/proj1/cs194-26-aao/#impl