Replicated Pasting 1

Replicated Filtering 1

We apply Perlin noise to the upper body of a query short walking clip. This operation is repeated over all matches to the query short walking clip. The Perlin noise is applied on the matches is the same frequency and amplitude as that applied on the query. Perlin noise adds texture to the motion such as nervousness, restlessness and weight shifting.

Matching parameters: LCSS, rotational parameterization, full-body area, no global position, no global rotation.

Replicated Filtering Example 1	Video Description (in order of appearance)
Full-size video	The query The best n-matches, from best to worst. Long pause The edited query The edited best n-matches, from best to worst. Long pause The edited query in context (i.e. with previous 50 and following 50 frames of animation). The edited best n-matches in context in context (i.e. with previous 50 and following 50 frames of animation).

Replicated Filtering 2 - Differentiation

Same as above, except that the frequency and amplitude of the Perlin noise varies randomly before being applied to each match. Each match gets to slightly different type of noise. This is useful when we want to try to differentiate each match from one another.

Matching parameters: LCSS, rotational parameterization, full-body area, no global position, no global rotation.

Pasting Example 2	Video Description (in order of appearance)
Full-size video	The query The best n-matches, from best to worst. Long pause The edited query The edited best n-matches, from best to worst. Long pause The edited query in context (i.e. with previous 50 and following 50 frames of animation). The edited best n-matches in context in context (i.e. with previous 50 and following 50 frames of animation).

Contextualised Replicated Filtering 3

This is an example of contextualised editing.

Same example as example 1 except that the intensity of the noise is controlled by the match strength. Therefore, the more similar the match is to the query, the more intense is the applied Perlin noise.

Contextualised Filtering Example 2	Video Description (in order of appearance)
Full-size video	The query The best n-matches, from best to worst. Long pause The edited query The edited best n-matches, from best to worst. Long pause The edited query in context (i.e. with previous 50 and following 50 frames of animation). The edited best n-matches in context in context (i.e. with previous 50 and following 50 frames of animation).