Postgres 12

By using db<>fiddle, you agree to license everything you submit by Creative Commons CC0.
CREATE TABLE images(id int, outline geography);
-- inserts the same polygon 1000 times
INSERT INTO images
SELECT id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geography
FROM generate_series(1,1000) id

CREATE TABLE

INSERT 0 1000

EXPLAIN (ANALYSE,BUFFERS)
SELECT
  i.id as id,
  ST_X(i.centroid) as lon,
  ST_Y(i.centroid) as lat
FROM (
  SELECT
    id,
    ST_Centroid(outline)::geometry AS centroid
  FROM images
) i;

QUERY PLAN

Seq Scan on images  (cost=0.00..17015.12 rows=6731 width=20) (actual time=113.743..365.799 rows=1000 loops=1)

  Buffers: shared hit=185 read=48 dirtied=2 written=1

Planning Time: 0.769 ms

Execution Time: 366.224 ms

EXPLAIN

EXPLAIN (ANALYSE,BUFFERS)
SELECT id,
  ST_X(centroid) as lon,
  ST_Y(centroid) as lat
FROM images
CROSS JOIN LATERAL ST_Centroid(outline::geometry) AS centroid 

QUERY PLAN

Nested Loop  (cost=1.25..289.84 rows=6731 width=20) (actual time=0.206..9.514 rows=1000 loops=1)

  Buffers: shared hit=53

  ->  Seq Scan on images  (cost=0.00..120.31 rows=6731 width=36) (actual time=0.026..0.316 rows=1000 loops=1)

        Buffers: shared hit=53

  ->  Function Scan on st_centroid centroid  (cost=1.25..1.26 rows=1 width=32) (actual time=0.008..0.008 rows=1 loops=1000)

Planning Time: 0.091 ms

Execution Time: 9.700 ms

EXPLAIN

-- Also possible like this, but considered to be a bad practice
EXPLAIN (ANALYSE,BUFFERS)
SELECT id,
  ST_X(centroid) as lon,
  ST_Y(centroid) as lat
FROM images, ST_Centroid(outline::geometry) AS centroid

QUERY PLAN

Nested Loop  (cost=1.25..289.84 rows=6731 width=20) (actual time=0.034..9.233 rows=1000 loops=1)

  Buffers: shared hit=53

  ->  Seq Scan on images  (cost=0.00..120.31 rows=6731 width=36) (actual time=0.008..0.281 rows=1000 loops=1)

        Buffers: shared hit=53

  ->  Function Scan on st_centroid centroid  (cost=1.25..1.26 rows=1 width=32) (actual time=0.008..0.008 rows=1 loops=1000)

Planning Time: 0.068 ms

Execution Time: 9.359 ms

EXPLAIN